1. Field of the Invention
The invention relates to an image processing method and an image processing apparatus, and more particularly to an image processing method and an image processing apparatus for re-mapping a local image dynamic response.
2. Description of the Related Art
An apparatus composed of a digital camera and other image processing apparatuses includes an image acquiring unit, an image processing unit, an image displaying unit and an image compressing unit. These units have different information amounts in processing with the same image source, and have image losses during the image processing procedure. Among these information losses, the gray-scale loss tends to eliminate the detail texture of the image. That is, the gray-scale loss tends to reduce the dynamic response range of the image processing apparatus. Especially, if the dynamic range of the gray-scale of the acquired image is very wide, for example, a flash is adopted or a back-lighted shot is performed, a detail brightness variation of the highlight or the shadow in the image generated at the last stage cannot be shown due to the insufficient information.
U.S. Pat. No. 6,256,424 (which is named '424 patent) discloses an image filter circuit utilizing the hardware architecture including an IIR-type low-frequency filter applied to generating an unsharp image for the dynamic range compression. The drawback of the '424 patent is that each edge of the processed image may become a pseudo-edge and cause distortion when the dynamic range compression ratio is large. U.S. Pat. No. 6,807,316 (which is named '316 patent) discloses an image processing method and an image processing apparatus using multiple low-frequency filters to generate an unsharp image applied to the dynamic range compression, and can thus eliminate the drawback of generating the pseudo-edge.
FIG. 1 shows the architecture of the image processing method and apparatus of the '316 patent. As shown in FIG. 1, the image processing apparatus utilize a first filtering process (Filtering Process 1) and a second filtering process (Filtering Process 2) to convert an input signal into a first unsharp image signal (Unsharp Image signal 1) and a second unsharp image signal (Unsharp Image signal 2), respectively. Then, a look-up-table synthesis (LUT Synthesis) is performed to synthesize the first unsharp image signal and the second unsharp image signal into a single image signal. Thereafter, a look-up-table computation (LUT Computation) is performed to generate a compensation signal. Finally, the input signal and the compensation signal are summated to generate an output signal.
The above-mentioned technology mainly utilizes a low-pass filter to separate a low-frequency signal from a high-frequency signal of an original image, and then reduces a gain of the low-frequency signal to compress the dynamic response range in the low-frequency area of the original image. Finally, the compressed low-frequency signal and the low-frequency area of the original image are synthesized such that the high-frequency signal (i.e., the detail texture) of the original image may be kept and the dynamic response range is also compressed within the range of the information amount that can be provided by an image processing unit. This technology has the drawback that the frequency response of all images is continuous. If some frequency simply serves as a dividing point for separating the low-frequency signal from the high-frequency signal, the final image may have the unsharp phenomenon as stated in the '424 patent. Thus, multiple dividing points corresponding to several stages of bands have to be used, as mentioned in the '316 patent, in order to try to reduce the unsharp phenomenon. In addition, this technology has another drawback that it is emphasized in the processing of the low-frequency dynamic response, and does not process the gains of the high-frequency signal in different highlight areas. Although this method can properly reduce the contrast ratio of the image with the high contrast ratio, it cannot provide a suitable gain for the high-frequency components of the image in the dark state and the highlight area. Thus, the final image becomes smooth but lacks the image fidelity.